feature: master presence timout for get space

Added timeout condition for the validation of master presence in
replicaset and for the master connection to the `utils.get_space`
method.

Closes #95

Author: GRISHNOV

CHANGELOG.md

@@ -8,6 +8,8 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 ## [Unreleased]
 
 ### Added
+* Added timeout condition for the validation of master presence in 
+  replicaset and for the master connection (#95).
 
 ### Changed
 

README.md

@@ -204,7 +204,8 @@ where:
 * `space_name` (`string`) - name of the space to insert an object
 * `tuple` / `object` (`table`) - tuple/object to insert
 * `opts`:
-  * `timeout` (`?number`) - `vshard.call` timeout (in seconds)
+  * `timeout` (`?number`) - `vshard.call` timeout and vshard master
+    discovery timeout (in seconds), default value is 2
   * `bucket_id` (`?number|cdata`) - bucket ID
   * `fields` (`?table`) - field names for getting only a subset of fields
   * `vshard_router` (`?string|table`) - Cartridge vshard group name or
@@ -263,7 +264,8 @@ where:
 * `space_name` (`string`) - name of the space to insert an object
 * `tuples` / `objects` (`table`) - array of tuples/objects to insert
 * `opts`:
-  * `timeout` (`?number`) - `vshard.call` timeout (in seconds)
+  * `timeout` (`?number`) - `vshard.call` timeout and vshard master
+    discovery timeout (in seconds), default value is 2
   * `fields` (`?table`) - field names for getting only a subset of fields
   * `stop_on_error` (`?boolean`) - stop on a first error and report error
     regarding the failed operation and error about what tuples were not
@@ -411,7 +413,8 @@ where:
 * `opts`:
   * `fields` (`?table`) - field names for getting only a subset of fields
   * `bucket_id` (`?number|cdata`) - bucket ID
-  * `timeout` (`?number`) - `vshard.call` timeout (in seconds)
+  * `timeout` (`?number`) - `vshard.call` timeout and vshard master
+    discovery timeout (in seconds), default value is 2
   * `mode` (`?string`, `read` or `write`) - if `write` is specified then `get` is
     performed on master, default value is `read`
   * `prefer_replica` (`?boolean`) - if `true` then the preferred target is one of
@@ -450,7 +453,8 @@ where:
 * `key` (`any`) - primary key value
 * `operations` (`table`) - update [operations](https://www.tarantool.io/en/doc/latest/reference/reference_lua/box_space/update/)
 * `opts`:
-  * `timeout` (`?number`) - `vshard.call` timeout (in seconds)
+  * `timeout` (`?number`) - `vshard.call` timeout and vshard master
+    discovery timeout (in seconds), default value is 2
   * `bucket_id` (`?number|cdata`) - bucket ID
   * `fields` (`?table`) - field names for getting only a subset of fields
   * `vshard_router` (`?string|table`) - Cartridge vshard group name or
@@ -485,7 +489,8 @@ where:
 * `space_name` (`string`) - name of the space
 * `key` (`any`) - primary key value
 * `opts`:
-  * `timeout` (`?number`) - `vshard.call` timeout (in seconds)
+  * `timeout` (`?number`) - `vshard.call` timeout and vshard master
+    discovery timeout (in seconds), default value is 2
   * `bucket_id` (`?number|cdata`) - bucket ID
   * `fields` (`?table`) - field names for getting only a subset of fields
   * `vshard_router` (`?string|table`) - Cartridge vshard group name or
@@ -522,7 +527,8 @@ where:
 * `space_name` (`string`) - name of the space
 * `tuple` / `object` (`table`) - tuple/object to insert or replace exist one
 * `opts`:
-  * `timeout` (`?number`) - `vshard.call` timeout (in seconds)
+  * `timeout` (`?number`) - `vshard.call` timeout and vshard master
+    discovery timeout (in seconds), default value is 2
   * `bucket_id` (`?number|cdata`) - bucket ID
   * `fields` (`?table`) - field names for getting only a subset of fields
   * `vshard_router` (`?string|table`) - Cartridge vshard group name or
@@ -581,7 +587,8 @@ where:
 * `space_name` (`string`) - name of the space to insert/replace an object
 * `tuples` / `objects` (`table`) - array of tuples/objects to insert
 * `opts`:
-  * `timeout` (`?number`) - `vshard.call` timeout (in seconds)
+  * `timeout` (`?number`) - `vshard.call` timeout and vshard master
+    discovery timeout (in seconds), default value is 2
   * `fields` (`?table`) - field names for getting only a subset of fields
   * `stop_on_error` (`?boolean`) - stop on a first error and report error
     regarding the failed operation and error about what tuples were not
@@ -730,7 +737,8 @@ where:
 * `tuple` / `object` (`table`) - tuple/object to insert if there is no existing tuple which matches the key fields
 * `operations` (`table`) - update [operations](https://www.tarantool.io/en/doc/latest/reference/reference_lua/box_space/update/) if there is an existing tuple which matches the key fields of tuple
 * `opts`:
-  * `timeout` (`?number`) - `vshard.call` timeout (in seconds)
+  * `timeout` (`?number`) - `vshard.call` timeout and vshard master
+    discovery timeout (in seconds), default value is 2
   * `bucket_id` (`?number|cdata`) - bucket ID
   * `fields` (`?table`) - field names for getting only a subset of fields
   * `vshard_router` (`?string|table`) - Cartridge vshard group name or
@@ -785,7 +793,8 @@ where:
    if there is tuple with duplicate key then existing tuple will
    be updated with update operations
 * `opts`:
-  * `timeout` (`?number`) - `vshard.call` timeout (in seconds)
+  * `timeout` (`?number`) - `vshard.call` timeout and vshard master
+    discovery timeout (in seconds), default value is 2
   * `fields` (`?table`) - field names for getting only a subset of fields
   * `stop_on_error` (`?boolean`) - stop on a first error and report error
     regarding the failed operation and error about what tuples were not
@@ -1108,7 +1117,8 @@ where:
 
 * `space_name` (`string`) - name of the space
 * `opts`:
-  * `timeout` (`?number`) - `vshard.call` timeout (in seconds)
+  * `timeout` (`?number`) - `vshard.call` timeout and vshard master
+    discovery timeout (in seconds), default value is 2
   * `vshard_router` (`?string|table`) - Cartridge vshard group name or
     vshard router instance. Set this parameter if your space is not
     a part of the default vshard cluster
@@ -1220,7 +1230,8 @@ where:
 * `opts`:
   * `yield_every` (`?number`) - number of tuples processed to yield after,
     `yield_every` should be > 0, default value is 100
-  * `timeout` (`?number`) - `vshard.call` timeout (in seconds), default value is 2
+  * `timeout` (`?number`) - `vshard.call` timeout and vshard master
+    discovery timeout (in seconds), default value is 2
   * `bucket_id` (`?number|cdata`) - bucket ID
   * `force_map_call` (`?boolean`) - if `true`
     then the map call is performed without any optimizations even,

crud/borders.lua

@@ -73,8 +73,7 @@ local function call_get_border_on_router(vshard_router, border_name, space_name,
         vshard_router = '?string|table',
     })
 
-    local replicasets = vshard_router:routeall()
-    local space, err = utils.get_space(space_name, replicasets)
+    local space, err = utils.get_space(space_name, vshard_router, opts.timeout)
     if err ~= nil then
         return nil, BorderError:new("An error occurred during the operation: %s", err), const.NEED_SCHEMA_RELOAD
     end
@@ -100,6 +99,10 @@ local function call_get_border_on_router(vshard_router, border_name, space_name,
     local cmp_key_parts = utils.merge_primary_key_parts(index.parts, primary_index.parts)
     local field_names = utils.enrich_field_names_with_cmp_key(opts.fields, cmp_key_parts, space:format())
 
+    local replicasets, err = vshard_router:routeall()
+    if err ~= nil then
+        return nil, BorderError:new("Failed to get router replicasets: %s", err)
+    end
     local call_opts = {
         mode = 'read',
         replicasets = replicasets,

crud/common/utils.lua

@@ -21,7 +21,7 @@ local FilterFieldsError = errors.new_class('FilterFieldsError', {capture_stack =
 local NotInitializedError = errors.new_class('NotInitialized')
 local StorageInfoError = errors.new_class('StorageInfoError')
 local VshardRouterError = errors.new_class('VshardRouterError', {capture_stack = false})
-local fiber_clock = require('fiber').clock
+local fiber = require('fiber')
 
 local utils = {}
 
@@ -96,8 +96,25 @@ function utils.format_replicaset_error(replicaset_uuid, msg, ...)
     )
 end
 
-function utils.get_space(space_name, replicasets)
-    local replicaset = select(2, next(replicasets))
+function utils.get_space(space_name, vshard_router, timeout)
+    local replicasets, replicaset
+    timeout = timeout or const.DEFAULT_VSHARD_CALL_TIMEOUT
+    local deadline = fiber.clock() + timeout
+    while (
+        -- Break if the deadline condition is exceeded.
+        -- Handling for deadline errors are below in the code.
+        fiber.clock() < deadline
+    ) do
+        -- Try to get master with timeout.
+        fiber.yield()
+        replicasets = vshard_router:routeall()
+        replicaset = select(2, next(replicasets))
+        if replicaset ~= nil and
+           replicaset.master ~= nil and
+           replicaset.master.conn.error == nil then
+            break
+        end
+    end
 
     if replicaset == nil then
         return nil, GetSpaceError:new(
@@ -119,13 +136,14 @@ function utils.get_space(space_name, replicasets)
              replicaset.uuid, replicaset.master.conn.error)
         return nil, GetSpaceError:new(error_msg)
     end
+
     local space = replicaset.master.conn.space[space_name]
 
     return space
 end
 
-function utils.get_space_format(space_name, replicasets)
-    local space, err = utils.get_space(space_name, replicasets)
+function utils.get_space_format(space_name, vshard_router)
+    local space, err = utils.get_space(space_name, vshard_router)
     if err ~= nil then
         return nil, GetSpaceFormatError:new("An error occurred during the operation: %s", err)
     end
@@ -664,7 +682,7 @@ function utils.cut_rows(rows, metadata, field_names)
 end
 
 local function flatten_obj(vshard_router, space_name, obj, skip_nullability_check)
-    local space_format, err = utils.get_space_format(space_name, vshard_router:routeall())
+    local space_format, err = utils.get_space_format(space_name, vshard_router)
     if err ~= nil then
         return nil, FlattenError:new("Failed to get space format: %s", err), const.NEED_SCHEMA_RELOAD
     end
@@ -835,9 +853,9 @@ function utils.storage_info(opts)
         end
     end
 
-    local deadline = fiber_clock() + timeout
+    local deadline = fiber.clock() + timeout
     for replica_uuid, future in pairs(futures_by_replicas) do
-        local wait_timeout = deadline - fiber_clock()
+        local wait_timeout = deadline - fiber.clock()
         if wait_timeout < 0 then
             wait_timeout = 0
         end

crud/count.lua

@@ -132,12 +132,7 @@ local function call_count_on_router(vshard_router, space_name, user_conditions,
         return nil, CountError:new("Failed to parse conditions: %s", err)
     end
 
-    local replicasets, err = vshard_router:routeall()
-    if err ~= nil then
-        return nil, CountError:new("Failed to get router replicasets: %s", err)
-    end
-
-    local space, err = utils.get_space(space_name, replicasets)
+    local space, err = utils.get_space(space_name, vshard_router, opts.timeout)
     if err ~= nil then
         return nil, CountError:new("An error occurred during the operation: %s", err), const.NEED_SCHEMA_RELOAD
     end
@@ -169,7 +164,10 @@ local function call_count_on_router(vshard_router, space_name, user_conditions,
     check_count_safety(space_name, plan, opts)
 
     -- set replicasets to count from
-    local replicasets_to_count = replicasets
+    local replicasets_to_count, err = vshard_router:routeall()
+    if err ~= nil then
+        return nil, CountError:new("Failed to get router replicasets: %s", err)
+    end
 
     -- Whether to call one storage replicaset or perform
     -- map-reduce?

crud/delete.lua

@@ -62,7 +62,7 @@ local function call_delete_on_router(vshard_router, space_name, key, opts)
         vshard_router = '?string|table',
     })
 
-    local space, err = utils.get_space(space_name, vshard_router:routeall())
+    local space, err = utils.get_space(space_name, vshard_router, opts.timeout)
     if err ~= nil then
         return nil, DeleteError:new("An error occurred during the operation: %s", err), const.NEED_SCHEMA_RELOAD
     end

crud/get.lua

@@ -65,7 +65,7 @@ local function call_get_on_router(vshard_router, space_name, key, opts)
         vshard_router = '?string|table',
     })
 
-    local space, err = utils.get_space(space_name, vshard_router:routeall())
+    local space, err = utils.get_space(space_name, vshard_router, opts.timeout)
     if err ~= nil then
         return nil, GetError:new("An error occurred during the operation: %s", err), const.NEED_SCHEMA_RELOAD
     end

crud/insert.lua

@@ -64,7 +64,7 @@ local function call_insert_on_router(vshard_router, space_name, original_tuple,
         vshard_router = '?string|table',
     })
 
-    local space, err = utils.get_space(space_name, vshard_router:routeall())
+    local space, err = utils.get_space(space_name, vshard_router, opts.timeout)
     if err ~= nil then
         return nil, InsertError:new("An error occurred during the operation: %s", err), const.NEED_SCHEMA_RELOAD
     end

crud/insert_many.lua

@@ -131,7 +131,7 @@ local function call_insert_many_on_router(vshard_router, space_name, original_tu
         vshard_router = '?string|table',
     })
 
-    local space, err = utils.get_space(space_name, vshard_router:routeall())
+    local space, err = utils.get_space(space_name, vshard_router, opts.timeout)
     if err ~= nil then
         return nil, {
             InsertManyError:new("An error occurred during the operation: %s", err)

crud/len.lua

@@ -64,7 +64,7 @@ function len.call(space_name, opts)
         return nil, LenError:new(err)
     end
 
-    local space, err = utils.get_space(space_name, vshard_router:routeall())
+    local space, err = utils.get_space(space_name, vshard_router, opts.timeout)
     if err ~= nil then
         return nil, LenError:new("An error occurred during the operation: %s", err)
     end